
summarize categorical data in a two-way table, compute joint, marginal and conditional relative frequencies, and use them to judge whether two variables are associated.
Instruction
Some data are categories, not numbers: grade level, yes or no, bus or car. To look for a relationship between two categorical variables, organize the data in a two-way table.
A survey (illustrative data). A school surveyed 120 students, 60 ninth graders and 60 twelfth graders, asking whether they have a paid job during the school year.
| Job | No job | Total | |
|---|---|---|---|
| Grade 9 | 12 | 48 | 60 |
| Grade 12 | 33 | 27 | 60 |
| Total | 45 | 75 | 120 |
The numbers inside the table are counts for each combination. The totals in the margins are the counts for each variable alone.
Three kinds of relative frequency. A relative frequency is a count divided by a total, usually written as a percent.
- Joint: a cell divided by the grand total. Ninth graders with jobs: 12 ÷ 120 = 10% of all students surveyed.
- Marginal: a margin total divided by the grand total. Students with jobs: 45 ÷ 120 = 37.5%.
- Conditional: a cell divided by its row or column total. Of ninth graders, 12 ÷ 60 = 20% have jobs. Of twelfth graders, 33 ÷ 60 = 55% have jobs.
Looking for association. Two categorical variables are associated if knowing one changes what you expect of the other. Compare the conditional relative frequencies across rows: 20% of ninth graders versus 55% of twelfth graders have jobs. If grade level and having a job were not associated, those percentages would be about the same, both near the overall 37.5%. They are far apart, so the data show an association.
Which total do you divide by? It depends on the question. “What percent of twelfth graders have jobs?” divides by the twelfth-grade row total: 55%. “What percent of students with jobs are twelfth graders?” divides by the job column total: 33 ÷ 45 ≈ 73.3%. These are different questions with different answers, and mixing them up is one of the most common errors in reading statistics in the news.
Making sense of it. The association is not surprising: older students are more likely to be old enough for many jobs and to have more flexible schedules. Notice what the table does not tell you: whether having a job affects grades, or why any individual student works. A two-way table describes a pattern in the group.
Try it. Survey your class (anonymously, with an option to decline) about two yes-or-no questions, or use the two-way table in this lesson’s reading. Build the two-way table, compute row percentages and write two sentences deciding whether the variables are associated.
📖 Read for this lesson
The reading opens over the lesson; close it to come back to exactly where you were. The checks below draw on it.
Formative check
Work through these before moving on. They are not graded — they tell you whether the standard below has landed yet.
What percent of students with jobs are ninth graders?
Of twelfth graders, % have jobs; of all students surveyed, % have jobs.
If 40% of ninth graders and 40% of twelfth graders had jobs, the data would show no association between grade and having a job.
Build a two-way table for two yes-or-no questions from your class (or the two-way table in this lesson’s reading). Compute the row percentages and write two sentences deciding whether the variables are associated, quoting the percentages you compared.
How confident are you that you can compute joint, marginal and conditional relative frequencies and use them to judge association?
Practice
Work these on paper or in your notebook, then open Check your answer. Aim for all of Fluency and Application; try at least one Challenge.
Printable version: this unit’s practice workbook (PDF)
Fluency
Build speed and accuracy with the core skill.
- A survey records two yes/no questions. Yes to both: 10; yes to A only: 55; yes to B only: 16; no to both: 20. Find P(A), P(A and B), P(A | B), and decide whether A and B look independent.
Check your answer
Answer: P(A) = 65/101 ≈ 0.644; P(A and B) = 10/101 ≈ 0.099; P(A | B) = 5/13 ≈ 0.385; not independent, since P(A | B) ≠ P(A).Build the two-way table first: rows A yes/no, columns B yes/no. - A survey records two yes/no questions. Yes to both: 53; yes to A only: 18; yes to B only: 59; no to both: 21. Find P(A), P(A and B), P(A | B), and decide whether A and B look independent.
Check your answer
Answer: P(A) = 71/151 ≈ 0.470; P(A and B) = 53/151 ≈ 0.351; P(A | B) = 53/112 ≈ 0.473; not independent in this data, since P(A | B) ≠ P(A); the difference is small, so in a wider population independence might be plausible.Build the two-way table first: rows A yes/no, columns B yes/no. - A survey records two yes/no questions. Yes to both: 29; yes to A only: 24; yes to B only: 14; no to both: 36. Find P(A), P(A and B), P(A | B), and decide whether A and B look independent.
Check your answer
Answer: P(A) = 53/103 ≈ 0.515; P(A and B) = 29/103 ≈ 0.282; P(A | B) = 29/43 ≈ 0.674; not independent, since P(A | B) ≠ P(A).Build the two-way table first: rows A yes/no, columns B yes/no. - A survey records two yes/no questions. Yes to both: 43; yes to A only: 14; yes to B only: 15; no to both: 18. Find P(A), P(A and B), P(A | B), and decide whether A and B look independent.
Check your answer
Answer: P(A) = 19/30 ≈ 0.633; P(A and B) = 43/90 ≈ 0.478; P(A | B) = 43/58 ≈ 0.741; not independent, since P(A | B) ≠ P(A).Build the two-way table first: rows A yes/no, columns B yes/no. - A table of 120 students: grade 9 — bus 18, car 30, walk 12; grade 12 — bus 10, car 42, walk 8. Find P(car | grade 12), P(grade 12 | car) and the joint relative frequency of grade 12 and car.
Check your answer
Answer: 70%, 58.3%, 35%Divide by the row total, the column total and the grand total. - In the same table, what percent of all students walk?
Check your answer
Answer: 16.7%Marginal: 20 ÷ 120.
Application
Use the skill in context. Show your reasoning.
- In a survey of 200 Missouri students (illustrative), 72 of 110 urban students and 38 of 90 rural students use a public library weekly. Compare conditional relative frequencies and decide whether the variables are associated.
Check your answer
Answer: Urban 65.5%, rural 42.2%: associated.Compare the row percentages. - In the same survey, what percent of all students use a library weekly?
Check your answer
Answer: 55%Marginal relative frequency: 110 ÷ 200. - A table shows 50 ninth graders and 70 twelfth graders. If there were no association, and 30% of each grade had a job, how many in each grade would have jobs?
Check your answer
Answer: 15 and 21Equal conditional percentages mean no association.
Challenge
Stretch problems. Expect to think before you write.
- Of 400 students, 25% are in group B, and 40% of group B answered yes to question A. How many answered yes and are in B? If A and B were independent, how many students overall would answer yes?
Check your answer
Answer: 40; 160Independence means P(A) = P(A | B) = 0.4.
Review
Keep earlier skills sharp.
- Solve the system: 6x + 5y = 31 and 3x − 5y = −22
Check your answer
Answer: (1, 5)Check the ordered pair in both equations. - Find the slope of the line through (−3, −6) and (0, 2).
Check your answer
Answer: m = 8/3m = (2 − (−6)) ÷ (0 − (−3)) = 8/3
A Real Two-Way Table: Schooling and Work, August 2026
BLS, The Employment Situation — August 2026 (released September 4, 2026), Table A-4, “Employment status of the civilian population 25 years and over by educational attainment,” seasonally adjusted, thousands of persons; retrieved September 23, 2026 · Public domain (works of the United States government)
Why this reading. The lesson’s two-way table is a survey of 120 students that somebody would have to run. Here is a real one, printed every month by the Bureau of Labor Statistics, covering about 148 million people. It is the same mathematics at a different scale, and it contains the exact confusion the lesson warns about — a row percentage and a column percentage that point in opposite directions — sitting in a table that gets quoted in the news.
Table A-4, August 2026
Civilian population aged 25 and over, seasonally adjusted, in thousands of people:
| Highest level of education | Employed | Unemployed | Civilian labor force |
|---|---|---|---|
| Less than a high school diploma | 7,513 | 371 | 7,885 |
| High school graduates, no college | 34,573 | 1,603 | 36,176 |
| Some college or associate degree | 34,740 | 1,319 | 36,059 |
| Bachelor’s degree and higher | 65,997 | 1,807 | 67,805 |
| Total | 142,823 | 5,100 | 147,925 |
A warning printed on the table itself: “Detail for the seasonally adjusted data shown in this table will not necessarily add to totals for those 25 years and over because of the independent seasonal adjustment of the various series.” You can see it happen. In the first row, 7,513 + 371 = 7,884, and the labor force is printed as 7,885. In the last row, 65,997 + 1,807 = 67,804 against a printed 67,805. Each series is adjusted for seasonality on its own, so the adjusted pieces need not add to the adjusted whole. This is not an error, and pretending it is not there is worse than explaining it in one sentence.
The three kinds of relative frequency, worked
Joint (a cell over the grand total). Unemployed people with a bachelor’s degree or higher: 1,807 ÷ 147,925 = 1.2% of everyone in this labor force.
Marginal (a margin over the grand total). Everyone with a bachelor’s degree or higher: 67,805 ÷ 147,925 = 45.8% of the labor force aged 25 and over.
Conditional, along a row (a cell over its row total). Of people with a bachelor’s degree or higher, 1,807 ÷ 67,805 = 2.7% are unemployed. That number has a name: it is the unemployment rate for that group, and it is what BLS prints beside the row.
| Group | Unemployment rate (row percentage) | Share of all unemployed (column percentage) |
|---|---|---|
| Less than a high school diploma | 4.7% | 7.3% |
| High school graduates, no college | 4.4% | 31.4% |
| Some college or associate degree | 3.7% | 25.9% |
| Bachelor’s degree and higher | 2.7% | 35.4% |
| Everyone 25 and over | 3.4% | 100% |
Check one of each: 371 ÷ 7,885 = 0.04705, which BLS prints as 4.7%. And 1,807 ÷ 5,100 = 0.3543, so 35.4%.
The two columns say opposite things, and both are true
Degree holders have the lowest unemployment rate of the four groups, 2.7%. Degree holders are also the largest single share of all unemployed people, 35.4% — more than any other group. There is no contradiction: they are 45.8% of the labor force, so even at the lowest rate they contribute a great many unemployed people.
A headline that says “more than a third of unemployed Americans have a college degree” is quoting the column. A headline that says “a degree halves your unemployment rate” is quoting the row. Dividing by the wrong total is the commonest mistake in reading statistics in the news, and this table is where it happens most often.
Is there an association?
Compare the conditional relative frequencies down the rows: 4.7%, 4.4%, 3.7%, 2.7%. If education and employment status were not associated, all four would sit near the overall 3.4%. They fall steadily instead, so the table shows an association, and a strong one: the rate at the top of the table is about 1.7 times the rate at the bottom.
What the table does not say
- Only people in the labor force appear. Anyone who is neither working nor actively looking is in none of these cells, and that share differs sharply by education.
- “Employed” counts anyone who did any paid work in the survey week, at any wage and any number of hours.
- It is one month of a survey, subject to sampling error and later revision.
- It describes groups, not people, and it shows association, not cause. What it cannot tell you is whether the schooling produced the difference; Lesson 4.4 is about exactly that.